client: drive App through felis-owned seams so its state machines are testable without winit #234

Merged

natsukium merged 4 commits from client/app-host-seam into main 2026-10-23 22:47:30 +09:00

Conversation 0 Commits 4 Files changed 15 +1523 -771

natsukium commented 2026-10-23 19:40:58 +09:00

Owner

Copy link

Closes #233.

The window-level state machines in felis-client (session switching, parked pushes, the exit ladder, the confirmation bar) could only be reached through a live winit event loop, so the 100-odd tests in tests.rs covered pure helpers only. This confines winit to a thin adapter and drives App from tests through felis-owned types. winit stays on 0.30.

Commits

1. Build and drive App without a winit event loop. Background tasks post through EventSink instead of EventLoopProxy. user_event becomes App::on_app_event, which records a window close in exit_requested for the adapter to forward. The window is built by App::new.

2. Route window effects through WindowHost. Every effect on an existing window (title, cursor, attention, IME area, fullscreen, decorations, backdrop, …) goes through one trait, which the winit window implements. App holds the window and the renderer separately, so a test can install a recording window without a GPU.

3. Hand App its input as felis types. window_event translates input into WindowInput (KeyPress after Option composition, WheelDelta, ImeInput, …), and App::on_window_input handles it. A bare modifier is its own PressedKey::Modifier, because the wire Key::Other would mix it up with dead keys, which the confirmation bar must still treat as an answer. Window events that act on the renderer (resize, scale factor, redraw, occlusion) stay in the adapter.

4. App-level tests for the rules these seams expose:

   • a reattach parked behind a failing switch runs when that switch fails
   • the title says "disconnected" while reconnecting
   • a BEL stamped with a stale connection generation does not flash the window
   • a bare modifier keeps the scrollback view and a typed key leaves it
   • confirmation bar: a bare modifier keeps it open, a dead or unbound key cancels, Y sends
   • the last parked push wins
   • a reattach to the current place dials nothing
   • a shell exit during a landing waits for that landing, then the exit ladder ends in a close
   • a refused roster fetch frees the next switch chord

   Each test in commit 4 fails under a hand mutation of the code it pins.

docs/reference/testing.md now mentions these window-less unit tests next to the frontend smoke.

Verification

• just check: green except deny, which fails on a yanked clipboard-win. Cargo.lock is unchanged by this PR, so main should hit the same failure.
• Windows cross-clippy: passes.
• just smoke-headless: passes.

Not included

• No regression test for a pipe/run chord on a session a chain re-pointed to. A successful landing needs a real carrier pair, which costs more than the test is worth here.

Closes #233. The window-level state machines in `felis-client` (session switching, parked pushes, the exit ladder, the confirmation bar) could only be reached through a live winit event loop, so the 100-odd tests in `tests.rs` covered pure helpers only. This confines winit to a thin adapter and drives `App` from tests through felis-owned types. winit stays on 0.30. ## Commits 1. **Build and drive `App` without a winit event loop.** Background tasks post through `EventSink` instead of `EventLoopProxy`. `user_event` becomes `App::on_app_event`, which records a window close in `exit_requested` for the adapter to forward. The window is built by `App::new`. 2. **Route window effects through `WindowHost`.** Every effect on an existing window (title, cursor, attention, IME area, fullscreen, decorations, backdrop, …) goes through one trait, which the winit window implements. `App` holds the window and the renderer separately, so a test can install a recording window without a GPU. 3. **Hand `App` its input as felis types.** `window_event` translates input into `WindowInput` (`KeyPress` after Option composition, `WheelDelta`, `ImeInput`, …), and `App::on_window_input` handles it. A bare modifier is its own `PressedKey::Modifier`, because the wire `Key::Other` would mix it up with dead keys, which the confirmation bar must still treat as an answer. Window events that act on the renderer (resize, scale factor, redraw, occlusion) stay in the adapter. 4. **App-level tests** for the rules these seams expose: - a reattach parked behind a failing switch runs when that switch fails - the title says "disconnected" while reconnecting - a BEL stamped with a stale connection generation does not flash the window - a bare modifier keeps the scrollback view and a typed key leaves it - confirmation bar: a bare modifier keeps it open, a dead or unbound key cancels, `Y` sends - the last parked push wins - a reattach to the current place dials nothing - a shell exit during a landing waits for that landing, then the exit ladder ends in a close - a refused roster fetch frees the next switch chord Each test in commit 4 fails under a hand mutation of the code it pins. `docs/reference/testing.md` now mentions these window-less unit tests next to the frontend smoke. ## Verification - `just check`: green except `deny`, which fails on a yanked `clipboard-win`. `Cargo.lock` is unchanged by this PR, so `main` should hit the same failure. - Windows cross-clippy: passes. - `just smoke-headless`: passes. ## Not included - No regression test for a pipe/run chord on a session a chain re-pointed to. A successful landing needs a real carrier pair, which costs more than the test is worth here.

natsukium added 4 commits 2026-10-23 19:40:58 +09:00

client: build and drive App without a winit event loop ... 32a0ceae3a

The window-level state machines (session switching, parked pushes,
the exit ladder) had no test at the App level because App reached its
loop through winit types a test cannot construct: an EventLoopProxy
for every event a background task posts, and an ActiveEventLoop for
every handler that closes the window.

Tasks now post through EventSink, a felis-owned closure the window
build backs with the proxy and a test with a channel. The user_event
handler is App::on_app_event, which records a close in exit_requested
for the winit adapter to forward, and the window's construction is
App::new, which a test calls with no warmup and no pump. A trait
object for the loop was rejected: the flag keeps every handler free
of a loop parameter, and only the adapter needs the loop.

The first test pins the parked-push contract: a reattach push that
arrives during a switch runs when that switch fails.

Assisted-by: Claude Code

client: route window effects through a WindowHost seam ... 614846692f

App reached its window through the same Option as its renderer, so
an App without a GPU could neither change nor be observed changing
its title, cursor or attention state, and every window effect was a
direct winit call scattered over the handlers.

WindowHost names every effect App has on an existing window, and the
winit window implements it. App now holds the window and the renderer
separately: a window build sets both in resumed, and a test sets only
a recorder. Creation stays on the concrete winit window, since the
renderer needs it and a test has no use for it.

Assisted-by: Claude Code

client: hand App its input as felis types ... aeb7cd07b2

The key, mouse, wheel and IME handlers took winit's own event types,
so a test could not press a key or move the pointer without a window,
and a winit release that reshapes those types would reach every input
handler.

The window_event adapter now translates what App handles into
WindowInput, a felis-owned enum, and App::on_window_input dispatches
it. Keys arrive as a KeyPress after Option composition; the bare
modifier the confirmation bar lets through is a PressedKey of its own,
since the wire Key folds it into Other together with the dead keys the
bar must still treat as an answer. The modifiers App tracks are the
felis set the wire and the keymap already use. Window events that act
on the renderer (resize, scale, redraw, occlusion) stay in the adapter.

Assisted-by: Claude Code

client: pin the window's switch, push and confirmation rules at App level ... 650e05f335

Each of these rules is a decision a pure helper already makes, but the
failures they guard against live in how App strings the decisions
together: a modifier that answers the confirmation bar, a parked push
lost to the one that replaced it, an exit ladder started beside a
landing still in flight, a refused roster fetch that leaves every later
switch chord queued behind it. Each test drives App through the events
and keys a window receives.

Assisted-by: Claude Code

natsukium force-pushed client/app-host-seam from 650e05f335 to 36c55eaf0e 2026-10-23 20:07:47 +09:00 Compare

natsukium force-pushed client/app-host-seam from 36c55eaf0e to c4fca227d2 2026-10-23 21:29:58 +09:00 Compare

natsukium scheduled this pull request to auto merge when all checks succeed 2026-10-23 21:50:34 +09:00

natsukium merged commit 631c33cd61 into main 2026-10-23 22:47:30 +09:00

natsukium deleted branch client/app-host-seam 2026-10-23 22:47:30 +09:00

natsukium referenced this pull request from a commit 2026-10-23 22:47:30 +09:00

Merge pull request 'client: drive App through felis-owned seams so its state machines are testable without winit' (#234) from client/app-host-seam into main

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels   

Area/ci

CI workflows (.forgejo/workflows/)

  

Area/cli

felis-cli crate, the felis front-door binary

  

Area/client

felis-client crate, the winit GUI frontend

  

Area/client-core

felis-client-core crate, shadow screen, config, connect_or_spawn_daemon

  

Area/daemon

felis-daemon crate, session pool, IPC server

  

Area/deps

Dependency bumps

  

Area/docs

Documentation tree when no narrower area fits

  

Area/grid

felis-grid crate, cells, scrollback, image store

  

Area/manual

User-facing docs: tutorials, how-to, user-facing reference pages

  

Area/nix

Nix flake, home-manager module, package definitions

  

Area/protocol

felis-protocol crate, felis.proto, frame layout, version constants

  

Area/pty

felis-pty crate, PTY backends

  

Area/render

felis-render-wgpu crate, GPU pipeline, atlas

  

Area/shaping

felis-shaping crate, swash shaper, font fallback

  

Area/skills

Agent skills (skills/felis, .agents/skills/)

  

Area/transport

felis-transport crate, framing, socket/pipe, SSH carrier

  

Area/vt

felis-vt crate, VT parser, dispatch

  

Compat/Breaking

Breaking change that won't be backward compatible

  

Kind/Bug

Something is not working

  

Kind/Documentation

Documentation changes

  

Kind/Enhancement

Improve existing functionality

  

Kind/Feature

New functionality

  

Kind/Performance

Throughput, latency, or memory cost

  

Kind/Security

This is security issue

  

Kind/Testing

Issue or pull request related to testing

  

perf-regression-accepted

Passes the instruction-count bench gate for an intended cost; state why in a PR comment

  

Reviewed

Confirmed

Issue has been confirmed

  

Reviewed

Duplicate

This issue or pull request already exists

  

Reviewed

Invalid

Invalid issue

  

Reviewed

Won't Fix

This issue won't be fixed

  

Status

Blocked

Something is blocking this issue or pull request

  

Status

Need More Info

Feedback is required to reproduce issue or to continue work

  

Status

Upstream

Waiting on a change in an upstream project

No labels

Area/ci

Area/cli

Area/client

Area/client-core

Area/daemon

Area/deps

Area/docs

Area/grid

Area/manual

Area/nix

Area/protocol

Area/pty

Area/render

Area/shaping

Area/skills

Area/transport

Area/vt

Compat/Breaking

Kind/Bug

Kind/Documentation

Kind/Enhancement

Kind/Feature

Kind/Performance

Kind/Security

Kind/Testing

perf-regression-accepted

Reviewed

Confirmed

Reviewed

Duplicate

Reviewed

Invalid

Reviewed

Won't Fix

Status

Blocked

Status

Need More Info

Status

Upstream

Milestone

Clear milestone

No items

No milestone

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

felis-terminal/felis!234

No description provided.